Scalable stability detection using logical hypercube

R. Friedman, S. Manor, K. Guo
Proceedings of the 18th IEEE Symposium on Reliable Distributed Systems  
This paper proposes to use a logical hypercube structure for detecting message stability in distributed systems. In particular, a stability detection protocol that uses such a superimposed l o gical structure i s p r esented, and its scalability is being compared with other known stability detection protocols. The main bene ts of the logical hypercube approach are scalability, faulttolerance, and refraining from overloading a single node or link in the system. These bene ts become evident by
more » ... h an analytic comparison and by simulations. Another important feature o f the logical hypercube approach is that the performance o f t h e p r otocol is in general not sensitive to the physical topology of the underlying (physical) network. Scalability: Each nodes need to communicate with only log(n) nodes. Performance: The logical hypercube structure guarantees that the maximal number of hops the stability information need to travel is at most log(n). Fault Tolerance: Hypercubes o er log(n) node distinct paths between every two nodes, therefore it can sustain up to log(n) failures. Regularity: Hypercubes have a very regular structure, and in our protocol every node plays exactly the same role. Thus, no node is more loaded than others. Also, code regularity tends to decrease the potential for software bugs in implementing the protocol.
doi:10.1109/reldis.1999.805089 dblp:conf/srds/FriedmanMG99 fatcat:afmikfispbd6hhjbgfc5ecl4su